INDEX

Explanations

politically charged terms or phrases

oai_token-act-pair · gpt-3.5-turbo

symbols or special characters in the text

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.5.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

enei

-0.70

 drain

-0.70

 drown

-0.67

sew

-0.67

 level

-0.67

 alley

-0.66

 ceilings

-0.65

 optimum

-0.64

 greenhouse

-0.64

 prom

-0.63

POSITIVE LOGITS

again

0.98

perhaps

0.95

whose

0.95

including

0.95

along

0.88

––

0.87

likely

0.87

pictured

0.84

particularly

0.83

via

0.83

Activations Density 0.215%

politically charged terms or phrases

symbols or special characters in the text

No Comments

No Known Activations